[BUG] Integration tests failing on Databricks-11.3 due to mixing of aggregations in HashAggregateExec and SortAggregateExec

[nartal1]

Describe the bug
Below tests are failing because there seems to be cpu aggregations(which we don't support) along with gpu aggregations. Needs more investigation.

• test_hash_groupby_collect_partial_replace_with_distinct_fallback
• test_groupby_std_variance_partial_replace_fallback
• test_hash_avg_nulls_partial_only
• test_groupby_std_variance
• test_groupby_std_variance_nulls

Reason for failure(Exec name changes but the reason is same) -

!Exec <SortAggregateExec> cannot run on GPU because mixing CPU and GPU aggregations is not supported. The data type of following expressions will be converted in GPU runtime: buf#216: Converted BinaryType to ArrayType(IntegerType,false); buf#218: Converted BinaryType to ArrayType(IntegerType,false)
!Exec <HashAggregateExec> cannot run on GPU because mixing CPU and GPU aggregations is not supported

Steps/Code to reproduce bug
Build and run integration tests on Databricks-11.3 environment.
Build steps - Pull in this PR locally if not merged - #7152
/home/ubuntu/spark-rapids$ ./jenkins/databricks/build.sh

Test steps: Modify the test.sh under jenkins/databrikcs to run a specific test.
Example: Comment all lines starting from export TEST_PARALLEL=${TEST_PARALLEL:-4} in the file and add below lines:

export TEST_PARALLEL=${TEST_PARALLEL:-1}
TEST_NAME=$1
bash /home/ubuntu/spark-rapids/integration_tests/run_pyspark_from_build.sh -k $TEST_NAME

Run Test - ./jenkins/databricks/test.sh test_hash_avg_nulls_partial_only

Expected behavior
All above tests should pass on Databricks-11.3 shim

Environment details (please complete the following information)
Databricks-11.3 runtime.

Additional context
Add any other context about the problem here.